Oncofusion

start / graph / mechanism-kmt2a-mllt1-engages-ptefb-through-the-enl-ahd-494-559

Mechanismstrong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.Q26b-mllt1-control-arm

KMT2A::MLLT1 engages P-TEFb through one 66-residue surface: the ENL AHD, residues 494-559, and L550E breaks it

also called MLL-ENL SEC engagement · ENL AHD AF4 P-TEFb · AEP recruitment by MLL-ENL

Everything this fusion does to speed up gene copying runs through a single short tail at the very end of the partner protein, just 66 amino acids long. Changing one amino acid in that tail stops the fusion from binding its partners and stops it causing leukaemia in cells.

The same thing, for a scientist

The elongation machinery reaches KMT2A::MLLT1 through the ENL Anc1-homology domain (AHD), mapped to residues 494-559 of the 559-residue protein (UniProt Q03111). That surface binds AF4-family proteins (AF4/AFF1, AF5q31/AFF4) and, alternately, DOT1L; a single L550E substitution abolishes both interactions and abolishes Hoxa9 upregulation and myeloid transformation by MLL-ENL. The complex the AHD recruits was purified from K562 as an endogenous five-member assembly of AF4, AF5q31, ENL, CDK9 and cyclin T1 eluting at ~0.8 MDa; neither that purification nor its AF5q31-directed counterpart recovered DOT1L or AF10. Native MLL-ENL in HB1119 forms an equivalent hybrid complex and co-localises with AF4 and AF5q31 at promoter-adjacent regions of HOXA9 and MEIS1, with negligible signal at ACTB and GAPDH. Two consequences are specific to this partner: the fusion does not incorporate wild-type ENL, and MLL-ENL-transformed cells are unaffected by wild-type ENL knockdown - unlike MLL-AF5q31- and MLL-AF6-transformed cells, which need it. Separately, transfecting MLL-ENL into HeLa increases the number of nuclear puncta containing AFF4 with ENL, with CDK9 and with ELL2, which has been proposed as a biophysical account of the same hijack.

What it looks like

No three-dimensional model is available for this protein, so the map above is the whole picture. That is itself informative: AlphaFold does not model chains this long, and the parts that have been crystallised are held under the entries for the individual domains.

Go deeper
The surface, to the residue

Structure-function analysis of MLL-ENL located the AF5q31 interaction to ENL residues 494-559 - the C-terminal 66 residues of a 559-residue protein, conserved with AF9 and with S. cerevisiae Anc1, hence AHD 1. The same AHD mediates DOT1L association. Mutations abolishing AF5q31 and DOT1L interaction, including the single point mutation L550E, caused failure to upregulate Hoxa9 and failure to transform myeloid progenitors 1.

The equivalent region of AF9 in MLL-AF9 is residues 502-568, and it behaves the same way 1. So the two YEATS-family KMT2A partners engage elongation through structurally equivalent C-terminal tails, and in both cases that tail is the transforming element.

The earlier two-hybrid and pulldown mapping had already put the AF4/AF5q31/AF10 binding domain of ENL in the C-terminus and noted that it 'coincided with the C-terminus that is essential for transformation by MLL-ENL' 2. The residue numbers come from the later work.

What is actually in the complex

Anti-AF4 immunoaffinity purification from K562 nuclear extract, after column chromatography, gave a mass-spectrometry hit list of AF5q31, ENL, CDK9 and cyclin T1; reciprocal IP confirmed all five proteins as one endogenous complex, and gel filtration put them co-eluting at an average mass of ~0.8 MDa 1. An AF5q31-directed purification gave the same thing.

A bounded negative from the same experiment, which matters: neither purification recovered DOT1L or AF10, both of which had been reported as ENL interactors 1. Read together with the mutual-exclusivity result in the same paper, that is consistent - ENL cannot hold AF4-family and DOT1L at the same time because both use the AHD.

Reciprocal IP in human leukaemia lines showed MLL-ENL, MLL-AF4 and MLL-AF5q31 each forming a comparable hybrid complex, while MLL-AF6 - a cytoplasmic partner - formed none, and wild-type MLL pulled down menin but no elongation components 1. The elongation complex and the MLL complex are separate biochemical entities that a subset of fusions welds together.

Where it sits on the genome

In HB1119, the reciprocal-positive KMT2A::MLLT1 line, ChIP showed MLL-ENL co-localising with AF4 and AF5q31 at the promoter-adjacent regions of HOXA9 and MEIS1, with AEP presence at non-target loci such as ACTB and GAPDH minimal or negligible 1. Co-localisation extended to CDKN1B and CDKN2C and into the transcribed bodies of HOXA9 and MEIS1, which is the pattern expected of an elongation function rather than an initiation one 1.

The murine MLL-ENL work adds a different target set - Eya1, Ms4a3, Hoxa10, Pim1, Mpo, Sox4, Nlrp3 down and Cdkn2c up on YEATS deletion 4 - so the HOXA9/MEIS1 axis is not the whole programme even in this fusion.

This fusion does not need wild-type ENL, and that is a problem for degrader strategies

Two observations in the same paper, and they belong together.

First, MLL-ENL did not form a complex with wild-type ENL in HB1119 1. Second, MLL-ENL-transformed cells were unaffected by wild-type ENL knockdown - their clonogenicity was unchanged - whereas MLL-AF5q31-transformed cells lost clonogenicity and Hoxa9 expression on the same knockdown, rescuable by exogenous human ENL, and MLL-AF6-transformed cells lost about 50% 1.

This cuts directly against the framing in which ENL is a general AML dependency. That framing comes from CRISPR and degradation work in lines that are mostly not t(11;19) 3. For a KMT2A::MLLT1 cell, wild-type ENL is not the thing keeping it alive; the fusion is. An ENL degrader or a YEATS antagonist acting on wild-type ENL would therefore be expected to work least well in the fusion that contains ENL - an inversion that is easy to miss when reading 'ENL-dependent leukaemia' as a category.

The partial counterweight is that a YEATS antagonist also occupies the groove of the YEATS domain inside the fusion, which 84.1% of patients retain, and HB1119 was more sensitive to SGC-iMLLT than KOPN-8 4. So the drug has two possible mechanisms in this disease and they have never been separated.

The condensate account of the same event

A separate laboratory reports that ENL, or its isolated IDR, is sufficient to initiate liquid droplet formation of the elongation complex in vitro, that AFF4 fluidises the droplets, and that ENL knockout disrupts heterotypic AFF4/cyclin T1 droplets in cells 5. Transfecting MLL-ENL into HeLa significantly increased the number of nuclear puncta containing AFF4 with ENL, AFF4 with CDK9 and AFF4 with ELL2 relative to wild-type ENL, which the authors offer as 'a biophysical mechanism for MLL-ENL fusion in leukemogenesis' 5.

That is the only published fusion-context condensate observation for this partner, and it should be weighted accordingly: transient overexpression in HeLa, puncta counting, and assays this project has documented as unable to distinguish phase separation from binding to spatially clustered sites. It is treated at length in the companion entry on ENL condensate biology.

Facts and figures
partnerMLLT1
uniprotQ03111
protein length AA559
accession resolutionQ03111 fetched from UniProt this session; primary gene name MLLT1, 559 aa, recommended name 'Protein ENL'
transformation dead point mutantL550E
complex mass MDa0.8
wild type enl required by mll enl cellsfalse
wild type enl required by mll af5q31 cellstrue
af9 equivalent ahdMLLT3 residues 502-568
domains
[
  {
    "name": "AHD (ANC1 homology domain)",
    "start": 494,
    "end": 559
  }
]
complex members
[
  "AFF1 (AF4)",
  "AFF4 (AF5q31)",
  "MLLT1 (ENL)",
  "CDK9",
  "CCNT1 (cyclin T1)"
]
not recovered in purification
[
  "DOT1L",
  "AF10 (MLLT10)"
]
chip targets hb1119
[
  "HOXA9",
  "MEIS1",
  "CDKN1B",
  "CDKN2C"
]
chip negative loci
[
  "ACTB",
  "GAPDH"
]
How this could bite us back
  • If wild-type ENL is dispensable in KMT2A::MLLT1 cells, an ENL degrader is a drug whose named target is not the dependency in the one fusion that contains it. A trial enriching for t(11;19) on the grounds that the partner is ENL would be selecting against its own mechanism.
  • The AHD is a 66-residue intrinsically disordered tail that folds on binding. Peptidomimetics against it have been made, but a disordered interface with micromolar partners is a hard target, and the same surface exists in wild-type ENL and wild-type AF9 throughout normal haematopoiesis.
  • Because the AHD binds AF4-family proteins and DOT1L through the same surface, a drug that blocks it removes both routes at once - which sounds good until one notices that it also removes the physiological elongation complex from every MLL target gene in normal cells.
  • A P-TEFb or CDK9 inhibitor reaches this fusion's mechanism but is a global transcription inhibitor with a narrow therapeutic index; the fusion-specific step is the tether, not the kinase.
Still open
  1. Does the reciprocal product MLLT1::KMT2A contain the AHD? The reciprocal allele would carry MLLT1 exon 1 plus the 3' half of KMT2A, so it should carry no AHD at all - which, if true, means the reciprocal product cannot compete for the elongation complex and the two control arms should be indistinguishable on this axis. Nobody has cloned it to check.
  2. Is L550E a usable separation-of-function tool in a human t(11;19) line? Every experiment on it is retroviral and murine, and knocking it into HB1119 would give the first endogenous test.
  3. Why does KMT2A::MLLT1 not incorporate wild-type ENL when the elongation complex has room for one ENL-family member per assembly? Is the fusion outcompeting it, or is the hybrid complex a different stoichiometry?
  4. Does an ENL degrader work in t(11;19) at all? The published degrader and YEATS-inhibitor packages do not report a KMT2A::MLLT1 line that is both YEATS-retaining and reciprocal-negative, and the wild-type-ENL dispensability result predicts they should underperform there.
Every source behind this entry (5)
  1. E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    ENL residues 494-559 (the AHD) mediate AF5q31 and DOT1L binding and are required for MLL-ENL transformation, with L550E sufficient to abolish both; the endogenous complex recovered from K562 is AF4/AF5q31/ENL/CDK9/cyclin T1 at ~0.8 MDa with no DOT1L or AF10; MLL-ENL co-localises with AF4 and AF5q31 at HOXA9 and MEIS1 in HB1119; and MLL-ENL neither incorporates nor requires wild-type ENL, unlike MLL-AF5q31.

    PMID 20153263DOI
    What this evidence cannot show

    The complex was purified from K562, a CML line with no KMT2A rearrangement, so its composition is that of the physiological complex rather than of the fusion's. The wild-type-ENL-dispensability result is shRNA clonogenicity in murine transformed cells, a proliferation readout over days, which would not detect a slow or in vivo requirement. Transformation assays use retroviral overexpression. Full text read from PMC2824033 via efetch db=pmc after fullTextXML returned HTTP 500.

    Evidence that argues against this (1)
    • An independent ENL immunopurification did recover DOT1L together with P-TEFb and reported H3K79 methylase activity in the same material, so the absence of DOT1L from the AF4-directed purification may be a property of the bait rather than of the biology. Mueller D et al. 2007
  2. E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    ENL interacts with AF4, AF5q31 and an AF10 fragment through a C-terminal region that coincides with the domain essential for transformation by MLL-ENL; ENL and AF4 co-localise in a nuclear speckled pattern and co-elute with Polycomb 3 on sizing columns; and ENL associates with histones H3 and H1 in a YEATS-dependent manner.

    PMID 15856011DOI
    What this evidence cannot show

    Read from abstract only: no PMCID exists for this record and all four retrieval routes failed, so no residue numbers, no quantification and no figure detail are available from this session. Two-hybrid and GST pulldown are interaction-prone assays. The nuclear-speckle colocalisation is of wild-type ENL with AF4, not of the fusion.

    Evidence that argues against this (1)
    • The later purification from the same field found that an AF4-directed immunoaffinity preparation did not recover AF10, one of the three interactors reported here, raising the possibility that some of these two-hybrid contacts are not stoichiometric in cells. Yokoyama A et al. 2010
  3. E3established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper

    Wild-type ENL is required for proliferation and disease maintenance in acute leukaemia cell lines identified by unbiased CRISPR screening and by targeted degradation, with acute ENL loss suppressing RNA polymerase II initiation and elongation genome-wide and an intact YEATS domain required for ENL-dependent growth.

    Transcription control by the ENL YEATS domain in acute leukaemia. · Erb MA, Scott TG, Li BE, et al. · Nature 2017
    PMID 28241139DOI
    What this evidence cannot show

    The screen was run in an MLL-AF4 line, and the dependency literature for wild-type ENL is dominated by non-t(11;19) backgrounds. Extending 'ENL is a leukaemia dependency' to KMT2A::MLLT1 is exactly the inference the wild-type-ENL knockdown result contradicts.

    Evidence that argues against this (1)
    • In HB1119, a t(11;19) line, MLL-ENL neither forms a complex with wild-type ENL nor requires it: shRNA knockdown left clonogenicity unchanged, while the same knockdown crippled MLL-AF5q31-transformed cells. Yokoyama A et al. 2010
  4. E4moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

    Loss of the ENL YEATS domain changes the MLL-ENL target programme (Eya1, Ms4a3, Hoxa10, Pim1, Mpo, Sox4 and Nlrp3 down, Cdkn2c up), and HB1119 cells show greater sensitivity to the ENL/AF9 YEATS inhibitor SGC-iMLLT at 2 uM than KOPN-8 cells, consistent with a contribution from the YEATS domain carried in the fusion.

    PMID 36435883DOI
    What this evidence cannot show

    HB1119 versus KOPN-8 differ in far more than YEATS status (patient age, lineage, KRAS status, reciprocal-allele status), so the comparison is not isogenic and cannot attribute the sensitivity difference to the fusion's YEATS domain. All human lines were sensitive at 20 uM, which the authors note may be cytotoxic. The murine target list is from one retroviral model.

    Evidence that argues against this (1)
    • In the same experiment MV4-11 (KMT2A::AFF1, no partner YEATS domain in the fusion at all) was also more sensitive at 2 uM, which the authors attribute to wild-type ENL - so the assay does not isolate the fusion-borne YEATS domain. Wan L et al. 2017
  5. E5moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

    ENL or its isolated IDR is sufficient to initiate liquid droplet formation of the super elongation complex in vitro, AFF4 fluidises those droplets, ENL knockout disrupts heterotypic AFF4/cyclin T1 droplets in cells, and transfecting MLL-ENL significantly increases the number of nuclear puncta containing AFF4 with ENL, with CDK9 and with ELL2 relative to wild-type ENL.

    PMID 32270036DOI
    What this evidence cannot show

    The MLL-ENL experiment is transient overexpression in HeLa, an epithelial line with no relevance to this disease, scored by counting colocalised puncta in >20 nuclei. The in vitro droplets use 20 uM purified IDR fusions with PEG-8000 crowding. 1,6-hexanediol and FRAP, the two assays used, cannot distinguish phase separation from binding to spatially clustered sites.

    Evidence that argues against this (1)
    • Only model-free calibrated half-bleaching discriminates liquid-liquid phase separation from binding to clustered sites; classical FRAP recovery and 1,6-hexanediol sensitivity, the assays used here, do not. Muzzopappa F et al. 2022
Connected to